Should You Take an ARM in 2026? The Break-Even Math When Fixed Rates Are 7.5%

The Spread Is the Whole Story

An ARM is worth considering only when the initial rate is enough below the fixed rate to pay you for taking the risk. Right now it is, by a wide margin.

From the MBA’s Weekly Applications Survey for the week ending October 9, 2026, average contract rates at 80% loan-to-value looked like this: conforming 30-year fixed 7.50% with 0.83 points; 30-year jumbo 7.49% with 0.70 points; 30-year FHA 7.19% with 1.21 points; 15-year fixed 6.87% with 0.90 points; and the 5/1 ARM at 6.54% with 1.13 points. Note the direction of travel within that week: the effective rate rose for the conforming, jumbo and 15-year products, held steady for FHA, and fell for the 5/1 ARM. Lenders are competing for ARM business.

Joel Kan, the MBA’s vice president and deputy chief economist, described the behavior plainly: “As a result, borrowers continue to rely more on ARM loans to help lower their monthly payments,” adding that “the ARM share at 10.5% was at its highest level since October 2025.”

ICE’s rate-lock data, which measures loans actually being locked rather than applications, shows the same shift and breaks it down usefully. For the week of September 18, 2026, ICE put the ARM share of refinance locks at 17.5% — the highest since it began tracking weekly in 2019 — and the ARM share of purchase locks at close to 9%, the second-highest weekly reading since 2022. Refinancers are reaching for ARMs harder than buyers are, which makes sense: a homeowner refinancing out of a payment they already cannot carry has a shorter effective horizon than someone buying a house to live in.

Two things to keep straight about that roughly one-point spread. It is not free money — it is the price the market is charging to take interest-rate risk off the lender’s books and put it on yours. And it is unusually wide by the standards of the past three years, which is precisely why the product is worth a look now when it was not worth one in 2024.

What an ARM Actually Is, Mechanically

Skip this section at your peril, because the three numbers in it determine your worst case, and almost nobody asks their lender for all three.

After the initial fixed period ends, your rate is rebuilt from scratch on each adjustment date out of two components. The Consumer Financial Protection Bureau defines the index as “an interest rate that fluctuates periodically based on general market conditions,” chosen by your lender when you apply and generally unchanged thereafter. The margin is “a number set by your lender when you apply for your loan” — a fixed number of percentage points added on top of the index, written into your loan agreement, and unchanged after closing. The CFPB’s formula is as simple as it sounds: “The fully indexed rate is equal to the margin plus the index.”

The margin is the part buyers systematically ignore, and the CFPB is explicit that it should not be: it varies a great deal between lenders and loans, and it can be negotiated exactly like the rate on a fixed-rate loan. Two ARMs quoted at the same teaser rate with margins 0.75 points apart are not the same loan, and the difference shows up every month for 25 years.

Then there are the caps, which are what stand between you and an unbounded payment. The CFPB describes three:

The initial adjustment cap limits the first change after the fixed period ends — commonly two or five percentage points, so the new rate “can’t be more than two (or five) percentage points higher or lower than the initial rate.”

The subsequent adjustment cap limits each later change, most commonly one or two percentage points.

The lifetime cap limits the total change over the life of the loan, most commonly five percentage points, so the rate “can never be more than five percentage points either higher or lower from the initial rate.” Some loans permit more, and some set a separate floor limiting how far the rate can fall.

The CFPB’s own shopping advice is the single most useful sentence in this entire subject: ask the lender to calculate the highest payment you could ever owe, and check it against your Loan Estimate and Truth-in-Lending disclosure. If a loan officer cannot produce that number on request, you are not dealing with a loan officer worth using.

The Five-Year Math on a Median-Priced House

Now the arithmetic. The national median existing-home price was $417,500 in September 2026, up 1.3% year over year (National Association of Realtors, October 13, 2026). Put 20% down and you are financing $334,000. Using the MBA’s October 9 contract rates, and ignoring points for a moment, our calculation of principal and interest comes out as follows.

At 7.50% fixed, the payment is $2,335 a month. At 6.54% on the ARM, it is $2,120. The difference is $215 a month, or $12,929 over the five-year fixed period.

That understates the advantage slightly, because a lower rate also means more of each payment goes to principal. After 60 payments the ARM balance is about $312,804 against $316,022 on the fixed loan — roughly $3,218 more equity. Add it to the payment saving and the five-year total advantage is about $16,147 on a $334,000 loan.

Two honest adjustments to that number. First, points: the MBA reported 1.13 points on the 5/1 ARM against 0.83 on the fixed, a difference of 0.30 points, which on $334,000 is roughly $1,000 of additional closing cost — call the net five-year advantage about $15,100. Second, this is principal and interest only. Taxes, insurance and any mortgage insurance are identical under both loans and do not change the comparison, though they do change whether you can afford either one; homeowners insurance in particular has stopped being a rounding error, which we covered in detail in what homeowners insurance actually costs in 2026.

Fifteen thousand dollars is a real number. It is also, crucially, money you keep only if you are out of the loan before the math reverses.

What the Reset Actually Costs — With Real Data

Here is where most ARM coverage goes soft, and where there is now excellent data.

Take the same $334,000 ARM at 6.54% and run it forward. At the first adjustment the balance is about $312,804 with 300 payments left. If the initial adjustment cap is two percentage points and the fully indexed rate is at or above the cap, the rate goes to 8.54% and the payment rises from $2,120 to about $2,527 — an increase of $407 a month, and $192 a month more than the 7.50% fixed payment you declined. A one-point adjustment to 7.54% puts you at about $2,320, essentially a wash against the fixed loan. And at the lifetime cap of five points — 11.54% — the payment is about $3,189, more than $1,000 a month above where you started. Those are our calculations, and they are the numbers your loan officer should be able to reproduce.

What makes this more than a hypothetical is that ICE is now tracking what resets do to real borrowers, and the figures are sobering. In its October 2026 Mortgage Monitor, ICE reported that about 180,000 ARMs are scheduled for their first reset in 2027, up from roughly 150,000 in 2026, with 155,000 more in 2028. The hardest-hit cohort is roughly 74,000 seven-year ARMs originated in 2020 — loans written at the bottom of the rate cycle, resetting near the top — with a median increase of about $1,066 a month, or 36%. Five-year ARMs written in 2022 fare better but still feel it: about 65,000 borrowers facing a median increase of roughly $430 a month, or 22%.

That is what a bad ARM outcome looks like in dollars: not a foreclosure wave, but a 22% to 36% payment increase arriving on a household budget built around the old number.

The reassuring half of ICE’s data is the scale. Andy Walden, ICE’s head of mortgage and housing market research, said that “ARMs are becoming more attractive to borrowers looking for relief from today’s higher fixed rates,” but “the overall market exposure to adjustable payments remains relatively limited.” There are 3.1 million active first-lien ARMs — the most in five and a half years — but they are just 5.6% of active mortgages, and only about 1.05 million have reached a first reset, the lowest count in more than 25 years. More than 90% of ARMs originated since 2022 are still inside their introductory fixed periods.

And the near-term sensitivity to the Fed is smaller than headlines imply. ICE calculated that a full 25-basis-point pass-through from the September hike raises the median affected ARM borrower’s payment by about $14 a month, or about $53 for more recently originated loans; a full percentage point of Fed tightening would mean about $56 a month. The danger in an ARM is not the Fed moving a quarter point. It is the step function at your own reset date.

The Question That Actually Decides It

Strip away everything else and an ARM is a bet on your own horizon. If you are reliably out of the loan — sold, or refinanced — before the first adjustment, you banked roughly $15,000 and took no risk that materialized. If you are still in it at year seven with a rate two points higher, you paid for someone else’s cheap money.

So: how long do people actually keep a house? The honest answer is much longer than they think when they sign. NAR’s 2025 Profile of Home Buyers and Sellers put median seller tenure at 11 years, an all-time high. The same report found first-time buyers at just 21% of the market, the lowest since the series began in 1981, with a median first-time buyer age of 40 and a median repeat buyer age of 62.

Read those numbers together and the base rate is unkind to a five-year ARM. The typical American who buys a house is in it for roughly a decade — twice the initial fixed period on a 5/1. If you take a five-year ARM on the assumption that you will be gone, you are betting against the median outcome, and you need a specific, concrete reason to believe you are the exception.

Good reasons exist, and they look like facts rather than intentions: a military assignment with a known rotation date, a medical residency or fellowship that ends in a specific year, a job on a visible relocation track, a house you are explicitly buying as a bridge, or a plan to sell a property whose proceeds will retire this loan. Bad reasons look like preferences: “we’ll probably want something bigger,” “we might move for work,” “five years is a long time.” If your exit is a hope rather than a date, price the loan as though you will be in it at reset — because statistically you will.

One structural note: longer initial periods are usually the better trade for an owner-occupant. A 7/6 or 10/6 ARM gives up some of the initial discount in exchange for a horizon that actually matches how long people stay. If the spread over the fixed rate is still meaningful at seven or ten years, that is generally the safer version of this bet than squeezing the last quarter point out of a five-year term.

The Refinance Assumption Is the Weak Link

Almost everyone who takes an ARM is quietly relying on a second assumption: that if rates fall, they will refinance into a fixed loan before the reset. This is the single most dangerous sentence in the transaction, because it requires three separate things to be true at once, and none of them is under your control.

Rates have to cooperate. The Fed is currently tightening, not easing. The 30-year fixed at 7.40% on October 15 (Freddie Mac) is near the top of this cycle, and Sam Khater, Freddie Mac’s chief economist, noted only that “the average 30-year fixed-rate mortgage remained unchanged this week.” Nobody — including anybody quoting a forecast at you — knows where the 30-year fixed sits in 2031.

You have to still qualify. A refinance is a full new underwrite: income, credit, debt-to-income, employment. Job loss, a business’s bad year, a new car loan, a medical event, or a divorce can all close that door at exactly the moment you need it. ICE reported the national delinquency rate at 3.53% in August 2026, up 14 basis points, with 574,000 loans 90 or more days past due but not in foreclosure, up 92,000 year over year — still below the 3.88% of August 2019, but moving the wrong way.

The house has to appraise. A refinance needs equity, and home prices are barely moving. The ICE Home Price Index was up 1.41% year over year in early September 2026, down from 1.44% in August, with single-family prices up 1.7% and condo prices down 0.9%; Seattle posted the largest annual decline among major markets at 3.1%. NAR has inventory at 1.60 million units and 4.8 months of supply in September 2026, up from 4.5 months a year earlier. In a flat-to-soft market, five years of amortization at 6.54% is most of the equity you will have — and if you bought with a small down payment in a metro where prices are falling, it may not be enough.

Refinance demand tells you how thin that door currently is: the MBA’s refinance index for the week ending October 9 was 56% below the same week a year earlier. The plan “we’ll just refinance” is a plan that depends on a market that does not currently exist.

ARM, Buydown, or 15-Year: Comparing the Three Discounts

An ARM is one of three ways to buy a lower rate, and they fail in different directions. Lay them side by side on the same $334,000 loan.

The ARM buys about 96 basis points — $215 a month on our numbers — and costs you nothing up front beyond a modest difference in points. What you pay is the reset risk, concentrated entirely on one date years from now. The discount is large, and so is the tail.

Discount points and temporary buydowns buy a smaller reduction for cash at closing, and the cash is gone whether or not you stay. The relevant question is the break-even month, and whether you will actually hold the loan past it; we worked through that math in should you buy down your mortgage rate in 2026. A seller-paid temporary buydown is the better version of this in today’s market, because someone else funds it — and with 4.8 months of supply and a median 33 days on market (NAR, September 2026), sellers are in a mood to fund things.

The 15-year fixed at 6.87% is the cheapest rate on the board that carries no rate risk at all. It is also the most expensive payment: about $2,978 a month on $334,000, roughly $643 more than the 30-year fixed. It is the right answer for a borrower whose constraint is total interest rather than monthly cash flow — which is to say, almost nobody who is attracted to an ARM.

The ordering that falls out of this is not complicated. If you can afford the fixed payment and your horizon is long, take the fixed loan and negotiate seller concessions toward a buydown. If you cannot afford the fixed payment, an ARM does not solve that problem — it postpones it, and the reset data above says the bill can arrive 22% to 36% larger. The ARM is for the borrower who can afford the fixed payment and has a dated, concrete reason to think the loan will be gone first.

Who Should Actually Consider One

An ARM is a defensible choice when most of the following are true at once, and a bad one when they are not.

You have a specific exit date, not an intention — a rotation, a residency end, a relocation, a bridge purchase. The initial fixed period comfortably exceeds that date, with margin for slippage; if your plan is to sell in year five, a 7/6 beats a 5/1. You could carry the fully capped payment if everything went wrong, which you verify by making the lender compute the maximum payment the CFPB tells you to ask for, then asking yourself whether you could pay it. You know your margin and your index, not just the teaser rate, and you shopped the margin. You are putting enough down that a flat market still leaves you able to sell or refinance at the reset. And the spread is wide enough to pay you for all of the above — near a full point, as it is now, rather than the quarter point ARMs often offer.

Conversely, skip it if you are stretching to qualify at the initial rate, if your exit is a preference rather than a date, if your income is variable or your employment uncertain, if you are buying at a high loan-to-value in a metro where prices are flat or falling, or if you cannot get a straight answer about the caps. And a note for refinancers specifically, since they are the group reaching for ARMs hardest at 17.5% of locks: refinancing out of a payment you cannot carry into an ARM you also will not be able to carry at reset is not a solution. It is a deferral with interest.

The Honest Bottom Line

The market is offering a genuinely large discount on adjustable-rate money right now — roughly 96 basis points, about $215 a month and $15,000 over five years on a median-priced house. That is not a gimmick and it is not 2006. The product is better regulated, the caps are disclosed, and ARMs are 5.6% of outstanding mortgages rather than a third of originations.

But the discount is payment for a risk that lands on one specific date, and the data on what that date costs is now unambiguous: a median increase of about $1,066 a month, or 36%, for the 2020 seven-year cohort who guessed that rates would be lower by now. They were not being reckless. They were being normal — and the median American stays in the house 11 years.

So the test is not whether you think rates will fall. It is whether you would be fine if they did not. Run the capped payment. Compare it to the fixed payment you are declining. If the capped number is one you could pay, the ARM is a reasonable trade and the spread is unusually generous. If it is a number that would break you, then what you are buying is not a cheaper mortgage — it is a cheaper first five years and a problem with a date on it.

And if the real issue is that the fixed payment at 7.50% does not work on your income, the honest answer is the unwelcome one: that is a signal about the price of the house, not the structure of the loan. For the broader picture of where rates sit and why, see mortgage rates just crossed 7%.